1 ; RUN: llc -march=mips -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=O32 -check-prefix=MIPS32 -check-prefix=ALL-BE %s 2 ; RUN: llc -march=mipsel -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=O32 -check-prefix=MIPS32 -check-prefix=ALL-LE %s 3 ; RUN: llc -march=mips64 -target-abi n32 -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=N32 -check-prefix=MIPS64 -check-prefix=ALL-BE %s 4 ; RUN: llc -march=mips64el -target-abi n32 -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=N32 -check-prefix=MIPS64 -check-prefix=ALL-LE %s 5 ; RUN: llc -march=mips64 -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=N64 -check-prefix=MIPS64 -check-prefix=ALL-BE %s 6 ; RUN: llc -march=mips64el -mattr=+msa,+fp64 < %s | FileCheck -check-prefix=ALL -check-prefix=N64 -check-prefix=MIPS64 -check-prefix=ALL-LE %s 7 8 @v4i8 = global <4 x i8> <i8 0, i8 0, i8 0, i8 0> 9 @v16i8 = global <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0> 10 @v8i16 = global <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0> 11 @v4i32 = global <4 x i32> <i32 0, i32 0, i32 0, i32 0> 12 @v2i64 = global <2 x i64> <i64 0, i64 0> 13 @i32 = global i32 0 14 @i64 = global i64 0 15 16 define void @const_v16i8() nounwind { 17 ; ALL-LABEL: const_v16i8: 18 19 store volatile <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, <16 x i8>*@v16i8 20 ; ALL: ldi.b [[R1:\$w[0-9]+]], 0 21 22 store volatile <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, <16 x i8>*@v16i8 23 ; ALL: ldi.b [[R1:\$w[0-9]+]], 1 24 25 store volatile <16 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 31>, <16 x i8>*@v16i8 26 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 27 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 28 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 29 ; ALL: ld.b [[R1:\$w[0-9]+]], 0([[G_PTR]]) 30 31 store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6>, <16 x i8>*@v16i8 32 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 33 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 34 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 35 ; ALL: ld.b [[R1:\$w[0-9]+]], 0([[G_PTR]]) 36 37 store volatile <16 x i8> <i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0, i8 1, i8 0>, <16 x i8>*@v16i8 38 ; ALL-BE: ldi.h [[R1:\$w[0-9]+]], 256 39 ; ALL-LE: ldi.h [[R1:\$w[0-9]+]], 1 40 41 store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4, i8 1, i8 2, i8 3, i8 4>, <16 x i8>*@v16i8 42 ; ALL-BE-DAG: lui [[R2:\$[0-9]+]], 258 43 ; ALL-LE-DAG: lui [[R2:\$[0-9]+]], 1027 44 ; ALL-BE-DAG: ori [[R2]], [[R2]], 772 45 ; ALL-LE-DAG: ori [[R2]], [[R2]], 513 46 ; ALL-DAG: fill.w [[R1:\$w[0-9]+]], [[R2]] 47 48 store volatile <16 x i8> <i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8>, <16 x i8>*@v16i8 49 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 50 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 51 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 52 ; ALL: ld.b [[R1:\$w[0-9]+]], 0([[G_PTR]]) 53 54 ret void 55 } 56 57 define void @const_v8i16() nounwind { 58 ; ALL-LABEL: const_v8i16: 59 60 store volatile <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, <8 x i16>*@v8i16 61 ; ALL: ldi.b [[R1:\$w[0-9]+]], 0 62 63 store volatile <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <8 x i16>*@v8i16 64 ; ALL: ldi.h [[R1:\$w[0-9]+]], 1 65 66 store volatile <8 x i16> <i16 1, i16 1, i16 1, i16 2, i16 1, i16 1, i16 1, i16 31>, <8 x i16>*@v8i16 67 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 68 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 69 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 70 ; ALL: ld.h [[R1:\$w[0-9]+]], 0([[G_PTR]]) 71 72 store volatile <8 x i16> <i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028, i16 1028>, <8 x i16>*@v8i16 73 ; ALL: ldi.b [[R1:\$w[0-9]+]], 4 74 75 store volatile <8 x i16> <i16 1, i16 2, i16 1, i16 2, i16 1, i16 2, i16 1, i16 2>, <8 x i16>*@v8i16 76 ; ALL-BE-DAG: lui [[R2:\$[0-9]+]], 1 77 ; ALL-LE-DAG: lui [[R2:\$[0-9]+]], 2 78 ; ALL-BE-DAG: ori [[R2]], [[R2]], 2 79 ; ALL-LE-DAG: ori [[R2]], [[R2]], 1 80 ; ALL-DAG: fill.w [[R1:\$w[0-9]+]], [[R2]] 81 82 store volatile <8 x i16> <i16 1, i16 2, i16 3, i16 4, i16 1, i16 2, i16 3, i16 4>, <8 x i16>*@v8i16 83 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 84 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 85 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 86 ; ALL: ld.h [[R1:\$w[0-9]+]], 0([[G_PTR]]) 87 88 ret void 89 } 90 91 define void @const_v4i32() nounwind { 92 ; ALL-LABEL: const_v4i32: 93 94 store volatile <4 x i32> <i32 0, i32 0, i32 0, i32 0>, <4 x i32>*@v4i32 95 ; ALL: ldi.b [[R1:\$w[0-9]+]], 0 96 97 store volatile <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32>*@v4i32 98 ; ALL: ldi.w [[R1:\$w[0-9]+]], 1 99 100 store volatile <4 x i32> <i32 1, i32 1, i32 1, i32 31>, <4 x i32>*@v4i32 101 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 102 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 103 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 104 ; ALL: ld.w [[R1:\$w[0-9]+]], 0([[G_PTR]]) 105 106 store volatile <4 x i32> <i32 16843009, i32 16843009, i32 16843009, i32 16843009>, <4 x i32>*@v4i32 107 ; ALL: ldi.b [[R1:\$w[0-9]+]], 1 108 109 store volatile <4 x i32> <i32 65537, i32 65537, i32 65537, i32 65537>, <4 x i32>*@v4i32 110 ; ALL: ldi.h [[R1:\$w[0-9]+]], 1 111 112 store volatile <4 x i32> <i32 1, i32 2, i32 1, i32 2>, <4 x i32>*@v4i32 113 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 114 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 115 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 116 ; ALL: ld.w [[R1:\$w[0-9]+]], 0([[G_PTR]]) 117 118 store volatile <4 x i32> <i32 3, i32 4, i32 5, i32 6>, <4 x i32>*@v4i32 119 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 120 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 121 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 122 ; ALL: ld.w [[R1:\$w[0-9]+]], 0([[G_PTR]]) 123 124 ret void 125 } 126 127 define void @const_v2i64() nounwind { 128 ; ALL-LABEL: const_v2i64: 129 130 store volatile <2 x i64> <i64 0, i64 0>, <2 x i64>*@v2i64 131 ; ALL: ldi.b [[R1:\$w[0-9]+]], 0 132 133 store volatile <2 x i64> <i64 72340172838076673, i64 72340172838076673>, <2 x i64>*@v2i64 134 ; ALL: ldi.b [[R1:\$w[0-9]+]], 1 135 136 store volatile <2 x i64> <i64 281479271743489, i64 281479271743489>, <2 x i64>*@v2i64 137 ; ALL: ldi.h [[R1:\$w[0-9]+]], 1 138 139 store volatile <2 x i64> <i64 4294967297, i64 4294967297>, <2 x i64>*@v2i64 140 ; ALL: ldi.w [[R1:\$w[0-9]+]], 1 141 142 store volatile <2 x i64> <i64 1, i64 1>, <2 x i64>*@v2i64 143 ; ALL: ldi.d [[R1:\$w[0-9]+]], 1 144 145 store volatile <2 x i64> <i64 1, i64 31>, <2 x i64>*@v2i64 146 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 147 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 148 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 149 ; MIPS32: ld.w [[R1:\$w[0-9]+]], 0([[G_PTR]]) 150 ; MIPS64: ld.d [[R1:\$w[0-9]+]], 0([[G_PTR]]) 151 152 store volatile <2 x i64> <i64 3, i64 4>, <2 x i64>*@v2i64 153 ; O32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %lo($ 154 ; N32: addiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 155 ; N64: daddiu [[G_PTR:\$[0-9]+]], {{.*}}, %got_ofst($ 156 ; MIPS32: ld.w [[R1:\$w[0-9]+]], 0([[G_PTR]]) 157 ; MIPS64: ld.d [[R1:\$w[0-9]+]], 0([[G_PTR]]) 158 159 ret void 160 } 161 162 define void @nonconst_v16i8(i8 signext %a, i8 signext %b, i8 signext %c, i8 signext %d, i8 signext %e, i8 signext %f, i8 signext %g, i8 signext %h) nounwind { 163 ; ALL-LABEL: nonconst_v16i8: 164 165 %1 = insertelement <16 x i8> undef, i8 %a, i32 0 166 %2 = insertelement <16 x i8> %1, i8 %b, i32 1 167 %3 = insertelement <16 x i8> %2, i8 %c, i32 2 168 %4 = insertelement <16 x i8> %3, i8 %d, i32 3 169 %5 = insertelement <16 x i8> %4, i8 %e, i32 4 170 %6 = insertelement <16 x i8> %5, i8 %f, i32 5 171 %7 = insertelement <16 x i8> %6, i8 %g, i32 6 172 %8 = insertelement <16 x i8> %7, i8 %h, i32 7 173 %9 = insertelement <16 x i8> %8, i8 %h, i32 8 174 %10 = insertelement <16 x i8> %9, i8 %h, i32 9 175 %11 = insertelement <16 x i8> %10, i8 %h, i32 10 176 %12 = insertelement <16 x i8> %11, i8 %h, i32 11 177 %13 = insertelement <16 x i8> %12, i8 %h, i32 12 178 %14 = insertelement <16 x i8> %13, i8 %h, i32 13 179 %15 = insertelement <16 x i8> %14, i8 %h, i32 14 180 %16 = insertelement <16 x i8> %15, i8 %h, i32 15 181 ; ALL-DAG: insert.b [[R1:\$w[0-9]+]][0], $4 182 ; ALL-DAG: insert.b [[R1]][1], $5 183 ; ALL-DAG: insert.b [[R1]][2], $6 184 ; ALL-DAG: insert.b [[R1]][3], $7 185 ; MIPS32-DAG: lw [[R2:\$[0-9]+]], 16($sp) 186 ; MIPS32-DAG: insert.b [[R1]][4], [[R2]] 187 ; MIPS64-DAG: insert.b [[R1]][4], $8 188 ; MIPS32-DAG: lw [[R3:\$[0-9]+]], 20($sp) 189 ; MIPS32-DAG: insert.b [[R1]][5], [[R3]] 190 ; MIPS64-DAG: insert.b [[R1]][5], $9 191 ; MIPS32-DAG: lw [[R4:\$[0-9]+]], 24($sp) 192 ; MIPS32-DAG: insert.b [[R1]][6], [[R4]] 193 ; MIPS64-DAG: insert.b [[R1]][6], $10 194 ; MIPS32-DAG: lw [[R5:\$[0-9]+]], 28($sp) 195 ; MIPS32-DAG: insert.b [[R1]][7], [[R5]] 196 ; MIPS64-DAG: insert.b [[R1]][7], [[R5:\$11]] 197 ; ALL-DAG: insert.b [[R1]][8], [[R5]] 198 ; ALL-DAG: insert.b [[R1]][9], [[R5]] 199 ; ALL-DAG: insert.b [[R1]][10], [[R5]] 200 ; ALL-DAG: insert.b [[R1]][11], [[R5]] 201 ; ALL-DAG: insert.b [[R1]][12], [[R5]] 202 ; ALL-DAG: insert.b [[R1]][13], [[R5]] 203 ; ALL-DAG: insert.b [[R1]][14], [[R5]] 204 ; ALL-DAG: insert.b [[R1]][15], [[R5]] 205 206 store volatile <16 x i8> %16, <16 x i8>*@v16i8 207 208 ret void 209 } 210 211 define void @nonconst_v8i16(i16 signext %a, i16 signext %b, i16 signext %c, i16 signext %d, i16 signext %e, i16 signext %f, i16 signext %g, i16 signext %h) nounwind { 212 ; ALL-LABEL: nonconst_v8i16: 213 214 %1 = insertelement <8 x i16> undef, i16 %a, i32 0 215 %2 = insertelement <8 x i16> %1, i16 %b, i32 1 216 %3 = insertelement <8 x i16> %2, i16 %c, i32 2 217 %4 = insertelement <8 x i16> %3, i16 %d, i32 3 218 %5 = insertelement <8 x i16> %4, i16 %e, i32 4 219 %6 = insertelement <8 x i16> %5, i16 %f, i32 5 220 %7 = insertelement <8 x i16> %6, i16 %g, i32 6 221 %8 = insertelement <8 x i16> %7, i16 %h, i32 7 222 ; ALL-DAG: insert.h [[R1:\$w[0-9]+]][0], $4 223 ; ALL-DAG: insert.h [[R1]][1], $5 224 ; ALL-DAG: insert.h [[R1]][2], $6 225 ; ALL-DAG: insert.h [[R1]][3], $7 226 ; MIPS32-DAG: lw [[R2:\$[0-9]+]], 16($sp) 227 ; MIPS32-DAG: insert.h [[R1]][4], [[R2]] 228 ; MIPS64-DAG: insert.h [[R1]][4], $8 229 ; MIPS32-DAG: lw [[R2:\$[0-9]+]], 20($sp) 230 ; MIPS32-DAG: insert.h [[R1]][5], [[R2]] 231 ; MIPS64-DAG: insert.h [[R1]][5], $9 232 ; MIPS32-DAG: lw [[R2:\$[0-9]+]], 24($sp) 233 ; MIPS32-DAG: insert.h [[R1]][6], [[R2]] 234 ; MIPS64-DAG: insert.h [[R1]][6], $10 235 ; MIPS32-DAG: lw [[R2:\$[0-9]+]], 28($sp) 236 ; MIPS32-DAG: insert.h [[R1]][7], [[R2]] 237 ; MIPS64-DAG: insert.h [[R1]][7], $11 238 239 store volatile <8 x i16> %8, <8 x i16>*@v8i16 240 241 ret void 242 } 243 244 define void @nonconst_v4i32(i32 signext %a, i32 signext %b, i32 signext %c, i32 signext %d) nounwind { 245 ; ALL-LABEL: nonconst_v4i32: 246 247 %1 = insertelement <4 x i32> undef, i32 %a, i32 0 248 %2 = insertelement <4 x i32> %1, i32 %b, i32 1 249 %3 = insertelement <4 x i32> %2, i32 %c, i32 2 250 %4 = insertelement <4 x i32> %3, i32 %d, i32 3 251 ; ALL: insert.w [[R1:\$w[0-9]+]][0], $4 252 ; ALL: insert.w [[R1]][1], $5 253 ; ALL: insert.w [[R1]][2], $6 254 ; ALL: insert.w [[R1]][3], $7 255 256 store volatile <4 x i32> %4, <4 x i32>*@v4i32 257 258 ret void 259 } 260 261 define void @nonconst_v2i64(i64 signext %a, i64 signext %b) nounwind { 262 ; ALL-LABEL: nonconst_v2i64: 263 264 %1 = insertelement <2 x i64> undef, i64 %a, i32 0 265 %2 = insertelement <2 x i64> %1, i64 %b, i32 1 266 ; MIPS32: insert.w [[R1:\$w[0-9]+]][0], $4 267 ; MIPS32: insert.w [[R1]][1], $5 268 ; MIPS32: insert.w [[R1]][2], $6 269 ; MIPS32: insert.w [[R1]][3], $7 270 ; MIPS64: insert.d [[R1:\$w[0-9]+]][0], $4 271 ; MIPS64: insert.d [[R1]][1], $5 272 273 store volatile <2 x i64> %2, <2 x i64>*@v2i64 274 275 ret void 276 } 277 278 define i32 @extract_sext_v16i8() nounwind { 279 ; ALL-LABEL: extract_sext_v16i8: 280 281 %1 = load <16 x i8>, <16 x i8>* @v16i8 282 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 283 284 %2 = add <16 x i8> %1, %1 285 ; ALL-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]] 286 287 %3 = extractelement <16 x i8> %2, i32 1 288 %4 = sext i8 %3 to i32 289 ; ALL-DAG: copy_s.b [[R3:\$[0-9]+]], [[R1]][1] 290 ; ALL-NOT: sll 291 ; ALL-NOT: sra 292 293 ret i32 %4 294 } 295 296 define i32 @extract_sext_v8i16() nounwind { 297 ; ALL-LABEL: extract_sext_v8i16: 298 299 %1 = load <8 x i16>, <8 x i16>* @v8i16 300 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 301 302 %2 = add <8 x i16> %1, %1 303 ; ALL-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]] 304 305 %3 = extractelement <8 x i16> %2, i32 1 306 %4 = sext i16 %3 to i32 307 ; ALL-DAG: copy_s.h [[R3:\$[0-9]+]], [[R1]][1] 308 ; ALL-NOT: sll 309 ; ALL-NOT: sra 310 311 ret i32 %4 312 } 313 314 define i32 @extract_sext_v4i32() nounwind { 315 ; ALL-LABEL: extract_sext_v4i32: 316 317 %1 = load <4 x i32>, <4 x i32>* @v4i32 318 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 319 320 %2 = add <4 x i32> %1, %1 321 ; ALL-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]] 322 323 %3 = extractelement <4 x i32> %2, i32 1 324 ; ALL-DAG: copy_s.w [[R3:\$[0-9]+]], [[R1]][1] 325 326 ret i32 %3 327 } 328 329 define i64 @extract_sext_v2i64() nounwind { 330 ; ALL-LABEL: extract_sext_v2i64: 331 332 %1 = load <2 x i64>, <2 x i64>* @v2i64 333 ; ALL-DAG: ld.d [[R1:\$w[0-9]+]], 334 335 %2 = add <2 x i64> %1, %1 336 ; ALL-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]] 337 338 %3 = extractelement <2 x i64> %2, i32 1 339 ; MIPS32-DAG: copy_s.w [[R3:\$[0-9]+]], [[R1]][2] 340 ; MIPS32-DAG: copy_s.w [[R4:\$[0-9]+]], [[R1]][3] 341 ; MIPS64-DAG: copy_s.d [[R3:\$[0-9]+]], [[R1]][1] 342 ; ALL-NOT: sll 343 ; ALL-NOT: sra 344 345 ret i64 %3 346 } 347 348 define i32 @extract_zext_v16i8() nounwind { 349 ; ALL-LABEL: extract_zext_v16i8: 350 351 %1 = load <16 x i8>, <16 x i8>* @v16i8 352 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 353 354 %2 = add <16 x i8> %1, %1 355 ; ALL-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]] 356 357 %3 = extractelement <16 x i8> %2, i32 1 358 %4 = zext i8 %3 to i32 359 ; ALL-DAG: copy_u.b [[R3:\$[0-9]+]], [[R1]][1] 360 ; ALL-NOT: andi 361 362 ret i32 %4 363 } 364 365 define i32 @extract_zext_v8i16() nounwind { 366 ; ALL-LABEL: extract_zext_v8i16: 367 368 %1 = load <8 x i16>, <8 x i16>* @v8i16 369 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 370 371 %2 = add <8 x i16> %1, %1 372 ; ALL-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]] 373 374 %3 = extractelement <8 x i16> %2, i32 1 375 %4 = zext i16 %3 to i32 376 ; ALL-DAG: copy_u.h [[R3:\$[0-9]+]], [[R1]][1] 377 ; ALL-NOT: andi 378 379 ret i32 %4 380 } 381 382 define i32 @extract_zext_v4i32() nounwind { 383 ; ALL-LABEL: extract_zext_v4i32: 384 385 %1 = load <4 x i32>, <4 x i32>* @v4i32 386 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 387 388 %2 = add <4 x i32> %1, %1 389 ; ALL-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]] 390 391 %3 = extractelement <4 x i32> %2, i32 1 392 ; ALL-DAG: copy_{{[su]}}.w [[R3:\$[0-9]+]], [[R1]][1] 393 394 ret i32 %3 395 } 396 397 define i64 @extract_zext_v2i64() nounwind { 398 ; ALL-LABEL: extract_zext_v2i64: 399 400 %1 = load <2 x i64>, <2 x i64>* @v2i64 401 ; ALL-DAG: ld.d [[R1:\$w[0-9]+]], 402 403 %2 = add <2 x i64> %1, %1 404 ; ALL-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]] 405 406 %3 = extractelement <2 x i64> %2, i32 1 407 ; MIPS32-DAG: copy_{{[su]}}.w [[R3:\$[0-9]+]], [[R1]][2] 408 ; MIPS32-DAG: copy_{{[su]}}.w [[R4:\$[0-9]+]], [[R1]][3] 409 ; MIPS64-DAG: copy_{{[su]}}.d [[R3:\$[0-9]+]], [[R1]][1] 410 ; ALL-NOT: andi 411 412 ret i64 %3 413 } 414 415 define i32 @extract_sext_v16i8_vidx() nounwind { 416 ; ALL-LABEL: extract_sext_v16i8_vidx: 417 418 %1 = load <16 x i8>, <16 x i8>* @v16i8 419 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v16i8)( 420 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v16i8)( 421 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v16i8)( 422 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 0([[PTR_V]]) 423 424 %2 = add <16 x i8> %1, %1 425 ; ALL-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]] 426 427 %3 = load i32, i32* @i32 428 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 429 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 430 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 431 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 432 433 %4 = extractelement <16 x i8> %2, i32 %3 434 %5 = sext i8 %4 to i32 435 ; ALL-DAG: splat.b $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 436 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 437 ; ALL-DAG: sra [[R6:\$[0-9]+]], [[R5]], 24 438 439 ret i32 %5 440 } 441 442 define i32 @extract_sext_v8i16_vidx() nounwind { 443 ; ALL-LABEL: extract_sext_v8i16_vidx: 444 445 %1 = load <8 x i16>, <8 x i16>* @v8i16 446 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v8i16)( 447 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v8i16)( 448 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v8i16)( 449 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 0([[PTR_V]]) 450 451 %2 = add <8 x i16> %1, %1 452 ; ALL-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]] 453 454 %3 = load i32, i32* @i32 455 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 456 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 457 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 458 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 459 460 %4 = extractelement <8 x i16> %2, i32 %3 461 %5 = sext i16 %4 to i32 462 ; ALL-DAG: splat.h $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 463 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 464 ; ALL-DAG: sra [[R6:\$[0-9]+]], [[R5]], 16 465 466 ret i32 %5 467 } 468 469 define i32 @extract_sext_v4i32_vidx() nounwind { 470 ; ALL-LABEL: extract_sext_v4i32_vidx: 471 472 %1 = load <4 x i32>, <4 x i32>* @v4i32 473 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v4i32)( 474 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v4i32)( 475 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v4i32)( 476 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 0([[PTR_V]]) 477 478 %2 = add <4 x i32> %1, %1 479 ; ALL-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]] 480 481 %3 = load i32, i32* @i32 482 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 483 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 484 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 485 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 486 487 %4 = extractelement <4 x i32> %2, i32 %3 488 ; ALL-DAG: splat.w $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 489 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 490 ; ALL-NOT: sra 491 492 ret i32 %4 493 } 494 495 define i64 @extract_sext_v2i64_vidx() nounwind { 496 ; ALL-LABEL: extract_sext_v2i64_vidx: 497 498 %1 = load <2 x i64>, <2 x i64>* @v2i64 499 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v2i64)( 500 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v2i64)( 501 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v2i64)( 502 ; ALL-DAG: ld.d [[R1:\$w[0-9]+]], 0([[PTR_V]]) 503 504 %2 = add <2 x i64> %1, %1 505 ; ALL-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]] 506 507 %3 = load i32, i32* @i32 508 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 509 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 510 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 511 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 512 513 %4 = extractelement <2 x i64> %2, i32 %3 514 ; MIPS32-DAG: splat.w $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 515 ; MIPS32-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 516 ; MIPS32-DAG: splat.w $w[[R4:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 517 ; MIPS32-DAG: mfc1 [[R6:\$[0-9]+]], $f[[R4]] 518 ; MIPS64-DAG: splat.d $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 519 ; MIPS64-DAG: dmfc1 [[R5:\$[0-9]+]], $f[[R3]] 520 ; ALL-NOT: sra 521 522 ret i64 %4 523 } 524 525 define i32 @extract_zext_v16i8_vidx() nounwind { 526 ; ALL-LABEL: extract_zext_v16i8_vidx: 527 528 %1 = load <16 x i8>, <16 x i8>* @v16i8 529 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v16i8)( 530 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v16i8)( 531 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v16i8)( 532 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 0([[PTR_V]]) 533 534 %2 = add <16 x i8> %1, %1 535 ; ALL-DAG: addv.b [[R2:\$w[0-9]+]], [[R1]], [[R1]] 536 537 %3 = load i32, i32* @i32 538 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 539 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 540 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 541 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 542 543 %4 = extractelement <16 x i8> %2, i32 %3 544 %5 = zext i8 %4 to i32 545 ; ALL-DAG: splat.b $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 546 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 547 ; ALL-DAG: srl [[R6:\$[0-9]+]], [[R5]], 24 548 549 ret i32 %5 550 } 551 552 define i32 @extract_zext_v8i16_vidx() nounwind { 553 ; ALL-LABEL: extract_zext_v8i16_vidx: 554 555 %1 = load <8 x i16>, <8 x i16>* @v8i16 556 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v8i16)( 557 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v8i16)( 558 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v8i16)( 559 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 0([[PTR_V]]) 560 561 %2 = add <8 x i16> %1, %1 562 ; ALL-DAG: addv.h [[R2:\$w[0-9]+]], [[R1]], [[R1]] 563 564 %3 = load i32, i32* @i32 565 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 566 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 567 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 568 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 569 570 %4 = extractelement <8 x i16> %2, i32 %3 571 %5 = zext i16 %4 to i32 572 ; ALL-DAG: splat.h $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 573 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 574 ; ALL-DAG: srl [[R6:\$[0-9]+]], [[R5]], 16 575 576 ret i32 %5 577 } 578 579 define i32 @extract_zext_v4i32_vidx() nounwind { 580 ; ALL-LABEL: extract_zext_v4i32_vidx: 581 582 %1 = load <4 x i32>, <4 x i32>* @v4i32 583 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v4i32)( 584 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v4i32)( 585 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v4i32)( 586 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 0([[PTR_V]]) 587 588 %2 = add <4 x i32> %1, %1 589 ; ALL-DAG: addv.w [[R2:\$w[0-9]+]], [[R1]], [[R1]] 590 591 %3 = load i32, i32* @i32 592 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 593 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 594 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 595 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 596 597 %4 = extractelement <4 x i32> %2, i32 %3 598 ; ALL-DAG: splat.w $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 599 ; ALL-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 600 ; ALL-NOT: srl 601 602 ret i32 %4 603 } 604 605 define i64 @extract_zext_v2i64_vidx() nounwind { 606 ; ALL-LABEL: extract_zext_v2i64_vidx: 607 608 %1 = load <2 x i64>, <2 x i64>* @v2i64 609 ; O32-DAG: lw [[PTR_V:\$[0-9]+]], %got(v2i64)( 610 ; N32-DAG: lw [[PTR_V:\$[0-9]+]], %got_disp(v2i64)( 611 ; N64-DAG: ld [[PTR_V:\$[0-9]+]], %got_disp(v2i64)( 612 ; ALL-DAG: ld.d [[R1:\$w[0-9]+]], 0([[PTR_V]]) 613 614 %2 = add <2 x i64> %1, %1 615 ; ALL-DAG: addv.d [[R2:\$w[0-9]+]], [[R1]], [[R1]] 616 617 %3 = load i32, i32* @i32 618 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 619 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 620 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 621 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 622 623 %4 = extractelement <2 x i64> %2, i32 %3 624 ; MIPS32-DAG: splat.w $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 625 ; MIPS32-DAG: mfc1 [[R5:\$[0-9]+]], $f[[R3]] 626 ; MIPS32-DAG: splat.w $w[[R4:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 627 ; MIPS32-DAG: mfc1 [[R6:\$[0-9]+]], $f[[R4]] 628 ; MIPS64-DAG: splat.d $w[[R3:[0-9]+]], [[R1]]{{\[}}[[IDX]]] 629 ; MIPS64-DAG: dmfc1 [[R5:\$[0-9]+]], $f[[R3]] 630 ; ALL-NOT: srl 631 632 ret i64 %4 633 } 634 635 define void @insert_v16i8(i32 signext %a) nounwind { 636 ; ALL-LABEL: insert_v16i8: 637 638 %1 = load <16 x i8>, <16 x i8>* @v16i8 639 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 640 641 %a2 = trunc i32 %a to i8 642 %a3 = sext i8 %a2 to i32 643 %a4 = trunc i32 %a3 to i8 644 ; ALL-NOT: andi 645 ; ALL-NOT: sra 646 647 %2 = insertelement <16 x i8> %1, i8 %a4, i32 1 648 ; ALL-DAG: insert.b [[R1]][1], $4 649 650 store <16 x i8> %2, <16 x i8>* @v16i8 651 ; ALL-DAG: st.b [[R1]] 652 653 ret void 654 } 655 656 define void @insert_v8i16(i32 signext %a) nounwind { 657 ; ALL-LABEL: insert_v8i16: 658 659 %1 = load <8 x i16>, <8 x i16>* @v8i16 660 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 661 662 %a2 = trunc i32 %a to i16 663 %a3 = sext i16 %a2 to i32 664 %a4 = trunc i32 %a3 to i16 665 ; ALL-NOT: andi 666 ; ALL-NOT: sra 667 668 %2 = insertelement <8 x i16> %1, i16 %a4, i32 1 669 ; ALL-DAG: insert.h [[R1]][1], $4 670 671 store <8 x i16> %2, <8 x i16>* @v8i16 672 ; ALL-DAG: st.h [[R1]] 673 674 ret void 675 } 676 677 define void @insert_v4i32(i32 signext %a) nounwind { 678 ; ALL-LABEL: insert_v4i32: 679 680 %1 = load <4 x i32>, <4 x i32>* @v4i32 681 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 682 683 ; ALL-NOT: andi 684 ; ALL-NOT: sra 685 686 %2 = insertelement <4 x i32> %1, i32 %a, i32 1 687 ; ALL-DAG: insert.w [[R1]][1], $4 688 689 store <4 x i32> %2, <4 x i32>* @v4i32 690 ; ALL-DAG: st.w [[R1]] 691 692 ret void 693 } 694 695 define void @insert_v2i64(i64 signext %a) nounwind { 696 ; ALL-LABEL: insert_v2i64: 697 698 %1 = load <2 x i64>, <2 x i64>* @v2i64 699 ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]], 700 ; MIPS64-DAG: ld.d [[R1:\$w[0-9]+]], 701 702 ; ALL-NOT: andi 703 ; ALL-NOT: sra 704 705 %2 = insertelement <2 x i64> %1, i64 %a, i32 1 706 ; MIPS32-DAG: insert.w [[R1]][2], $4 707 ; MIPS32-DAG: insert.w [[R1]][3], $5 708 ; MIPS64-DAG: insert.d [[R1]][1], $4 709 710 store <2 x i64> %2, <2 x i64>* @v2i64 711 ; MIPS32-DAG: st.w [[R1]] 712 ; MIPS64-DAG: st.d [[R1]] 713 714 ret void 715 } 716 717 define void @insert_v16i8_vidx(i32 signext %a) nounwind { 718 ; ALL-LABEL: insert_v16i8_vidx: 719 720 %1 = load <16 x i8>, <16 x i8>* @v16i8 721 ; ALL-DAG: ld.b [[R1:\$w[0-9]+]], 722 723 %2 = load i32, i32* @i32 724 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 725 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 726 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 727 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 728 729 %a2 = trunc i32 %a to i8 730 %a3 = sext i8 %a2 to i32 731 %a4 = trunc i32 %a3 to i8 732 ; ALL-NOT: andi 733 ; ALL-NOT: sra 734 735 %3 = insertelement <16 x i8> %1, i8 %a4, i32 %2 736 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[IDX]]] 737 ; ALL-DAG: insert.b [[R1]][0], $4 738 ; O32-DAG: neg [[NIDX:\$[0-9]+]], [[IDX]] 739 ; N32-DAG: neg [[NIDX:\$[0-9]+]], [[IDX]] 740 ; N64-DAG: dneg [[NIDX:\$[0-9]+]], [[IDX]] 741 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 742 743 store <16 x i8> %3, <16 x i8>* @v16i8 744 ; ALL-DAG: st.b [[R1]] 745 746 ret void 747 } 748 749 define void @insert_v8i16_vidx(i32 signext %a) nounwind { 750 ; ALL-LABEL: insert_v8i16_vidx: 751 752 %1 = load <8 x i16>, <8 x i16>* @v8i16 753 ; ALL-DAG: ld.h [[R1:\$w[0-9]+]], 754 755 %2 = load i32, i32* @i32 756 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 757 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 758 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 759 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 760 761 %a2 = trunc i32 %a to i16 762 %a3 = sext i16 %a2 to i32 763 %a4 = trunc i32 %a3 to i16 764 ; ALL-NOT: andi 765 ; ALL-NOT: sra 766 767 %3 = insertelement <8 x i16> %1, i16 %a4, i32 %2 768 ; ALL-DAG: sll [[BIDX:\$[0-9]+]], [[IDX]], 1 769 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[BIDX]]] 770 ; ALL-DAG: insert.h [[R1]][0], $4 771 ; O32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 772 ; N32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 773 ; N64-DAG: dneg [[NIDX:\$[0-9]+]], [[BIDX]] 774 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 775 776 store <8 x i16> %3, <8 x i16>* @v8i16 777 ; ALL-DAG: st.h [[R1]] 778 779 ret void 780 } 781 782 define void @insert_v4i32_vidx(i32 signext %a) nounwind { 783 ; ALL-LABEL: insert_v4i32_vidx: 784 785 %1 = load <4 x i32>, <4 x i32>* @v4i32 786 ; ALL-DAG: ld.w [[R1:\$w[0-9]+]], 787 788 %2 = load i32, i32* @i32 789 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 790 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 791 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 792 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 793 794 ; ALL-NOT: andi 795 ; ALL-NOT: sra 796 797 %3 = insertelement <4 x i32> %1, i32 %a, i32 %2 798 ; ALL-DAG: sll [[BIDX:\$[0-9]+]], [[IDX]], 2 799 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[BIDX]]] 800 ; ALL-DAG: insert.w [[R1]][0], $4 801 ; O32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 802 ; N32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 803 ; N64-DAG: dneg [[NIDX:\$[0-9]+]], [[BIDX]] 804 ; ALL-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 805 806 store <4 x i32> %3, <4 x i32>* @v4i32 807 ; ALL-DAG: st.w [[R1]] 808 809 ret void 810 } 811 812 define void @insert_v2i64_vidx(i64 signext %a) nounwind { 813 ; ALL-LABEL: insert_v2i64_vidx: 814 815 %1 = load <2 x i64>, <2 x i64>* @v2i64 816 ; MIPS32-DAG: ld.w [[R1:\$w[0-9]+]], 817 ; MIPS64-DAG: ld.d [[R1:\$w[0-9]+]], 818 819 %2 = load i32, i32* @i32 820 ; O32-DAG: lw [[PTR_I:\$[0-9]+]], %got(i32)( 821 ; N32-DAG: lw [[PTR_I:\$[0-9]+]], %got_disp(i32)( 822 ; N64-DAG: ld [[PTR_I:\$[0-9]+]], %got_disp(i32)( 823 ; ALL-DAG: lw [[IDX:\$[0-9]+]], 0([[PTR_I]]) 824 825 ; ALL-NOT: andi 826 ; ALL-NOT: sra 827 828 %3 = insertelement <2 x i64> %1, i64 %a, i32 %2 829 ; TODO: This code could be a lot better but it works. The legalizer splits 830 ; 64-bit inserts into two 32-bit inserts because there is no i64 type on 831 ; MIPS32. The obvious optimisation is to perform both insert.w's at once while 832 ; the vector is rotated. 833 ; MIPS32-DAG: sll [[BIDX:\$[0-9]+]], [[IDX]], 2 834 ; MIPS32-DAG: sld.b [[R1]], [[R1]]{{\[}}[[BIDX]]] 835 ; MIPS32-DAG: insert.w [[R1]][0], $4 836 ; MIPS32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 837 ; MIPS32-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 838 ; MIPS32-DAG: addiu [[IDX2:\$[0-9]+]], [[IDX]], 1 839 ; MIPS32-DAG: sll [[BIDX:\$[0-9]+]], [[IDX2]], 2 840 ; MIPS32-DAG: sld.b [[R1]], [[R1]]{{\[}}[[BIDX]]] 841 ; MIPS32-DAG: insert.w [[R1]][0], $5 842 ; MIPS32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 843 ; MIPS32-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 844 845 ; MIPS64-DAG: sll [[BIDX:\$[0-9]+]], [[IDX]], 3 846 ; MIPS64-DAG: sld.b [[R1]], [[R1]]{{\[}}[[BIDX]]] 847 ; MIPS64-DAG: insert.d [[R1]][0], $4 848 ; N32-DAG: neg [[NIDX:\$[0-9]+]], [[BIDX]] 849 ; N64-DAG: dneg [[NIDX:\$[0-9]+]], [[BIDX]] 850 ; MIPS64-DAG: sld.b [[R1]], [[R1]]{{\[}}[[NIDX]]] 851 852 store <2 x i64> %3, <2 x i64>* @v2i64 853 ; MIPS32-DAG: st.w [[R1]] 854 ; MIPS64-DAG: st.d [[R1]] 855 856 ret void 857 } 858 859 define void @truncstore() nounwind { 860 ; ALL-LABEL: truncstore: 861 862 store volatile <4 x i8> <i8 -1, i8 -1, i8 -1, i8 -1>, <4 x i8>*@v4i8 863 ; TODO: What code should be emitted? 864 865 ret void 866 } 867
